2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Sunday, December 13, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Molecular Testing, Omics Technologies, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Drug Development, Therapy Sequence, Biological Therapies, Antibody Therapy, Transplantation, Checkpoint Inhibitor, Immunotherapy, Non-Biological Therapies, Chemotherapy, Adverse Events, Artificial Intelligence (AI), Deep Learning, Machine Learning, Biological Processes, Apoptosis, Pathogenesis, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Practice Models, Value, Diseases, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, CMML, Clinical Research, Epidemiology, Health Outcomes Research, Health Disparities Research, Real-World Evidence, Survivorship, Study Population, Human, Adult
PAPER

3437 Recurrent non-canonical proteoforms in acute myeloid leukemia identified by integrative proteogenomics

Laura Katharina Schmalbrock, MD, Asher Preska Steinberg, PhD2, Katarzyna Kulej, PhD3, Jinxin Zhang, PhD4, Gabriella Casalena, PhD5, Andrew McPherson, PhD6, Alex Kentsis, MD, PhD7
1 Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
2 Computational Oncology Program, Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
3 Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
4 Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
5 Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
6 Computational Oncology Program, Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York, United States
7 Departments of Pediatrics, Pharmacology, and Physiology & Biophysics, Weill Medical College of Cornell University, New York, New York, United States, Molecular Pharmacology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center,, New York, New York, United States, Tow Center for Developmental Oncology, Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York, United States

Background: The reference proteome incompletely represents the proteins translated in cancer. Genomic variation, aberrant transcription or splicing, and noncanonical translation can generate proteoforms absent from curated databases such as UniProtKB/Swiss-Prot, including microproteins translated from small open reading frames (smORFs) of fewer than 100 amino acids. This poorly mapped “dark proteome” may contain previously unrecognized regulators of leukemogenesis and a reservoir of tumor-selective therapeutic targets. However, its extent, recurrence, and molecular features have not been systematically defined in acute myeloid leukemia (AML).

Methods: We developed a cohort-informed proteogenomic strategy to interrogate the AML dark proteome in published paired Illumina RNA-sequencing and mass-spectrometry (MS) datasets from 123 AML samples and MS data from 13 healthy CD34+ controls. Using ProteomeGenerator2, we performed de novo transcriptome assembly with StringTie and ORF prediction with TransDecoder, explicitly retaining non-canonical ORFs and smORFs. The resulting patient-specific sequence database was searched against raw MS data in FragPipe. Candidate reference-unannotated sequences were prioritized based on unique high-quality peptide support, absence from healthy CD34+ controls, recurrence across AML samples, and without sequence homology to annotated proteins by DIAMOND BLASTP. Orthogonal support was sought in deep, fractionated, multiprotease DIA proteomics of five AML cell lines and two CD34+ controls, using trypsin, LysN, GluC, chymotrypsin, and AspN digestion. Protein features were evaluated with InterPro/Pfam and AlphaFold 3.

Results: We identified 13,265 Swiss-Prot-annotated protein isoforms, representing 9,682 protein groups, together with 5,849 protein sequences not represented in Swiss-Prot. Swiss-Prot annotated proteoforms were commonly shared between AML and healthy CD34+ samples: 6,576 (49.6%) were detected in both. In contrast, Swiss-Prot unannotated candidates were strongly enriched in AML, with only 537 detected in both AML and CD34+ controls. Proteoforms detected in only one AML sample were disproportionately reference-unannotated (3,127 of 4,528), whereas proteoforms detected across all AML samples were almost exclusively Swiss-Prot annotated (847 of 860).

After excluding sequences detected in CD34+ controls and applying DIAMOND BLASTP filtering, we identified 1,987 candidate protein sequences without close similarity to annotated proteins. Thirty-nine were recurrently detected in more than 10% of AML samples, defining a set of recurrent AML-associated reference-unannotated candidates. Thirty-three of these proteins were encoded by smORFs shorter than 100 amino acids, and all were supported by at least one unique, high-quality peptide.

Independent deep proteomic analysis identified 14 AML-enriched, reference-unannotated candidates shared between primary AML samples and AML cell lines; 13 encoded predicted microproteins. Across the combined set of 53 recurrent or independently replicated candidates, 26 lacked identifiable InterPro/Pfam annotations, 15 contained predicted intrinsically disordered regions, five contained transmembrane regions, four harbored recognizable domains, and three encoded predicted signal peptides. AlphaFold 3 predicted confident structures for 19 candidates (mean pLDDT >70), predominantly alpha-helical, whereas 34 had lower-confidence structural predictions, including 11 with predicted intrinsically disordered regions.

Conclusions: These results provide a cohort-scale map of the AML dark proteome and identify recurrent, AML-enriched proteoforms not represented in standard reference databases. The predominance of small, poorly annotated, and frequently disordered sequences suggests that the unexplored translated genome may encode a distinct class of cancer-associated proteins that is systematically underdetected by conventional proteomics.

Our findings also define key requirements for improved dark-proteome discovery. Long-read RNA sequencing, ribosome profiling, targeted proteomics, and perturbational studies will be important next steps to distinguish bona fide translated and functional proteoforms from transcriptional or analytical candidates. Together, these methods will enable systematic discovery of previously undetected leukemia proteins and prioritization of tumor-specific, candidate proteoforms for mechanistic and therapeutic development.

Laura Katharina Schmalbrock, MD: No relevant conflicts of interest to declare.